[Beowulf] Keeping the Athlon MP cluster limping along
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Robert G. Brown rgb at phy.duke.eduThu Dec 9 11:53:26 PST 2004
- Previous message: [Beowulf] Keeping the Athlon MP cluster limping along
- Next message: [Beowulf] Keeping the Athlon MP cluster limping along
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Wed, 8 Dec 2004, David Mathog wrote: > It's official, the Tyan S2466 nodes get "biggest PITA award" > for systems that I've used. The two nodes that were crashing > frequently had their power supplies replaced and then they > were stable for a couple of months. Now they've both become > unstable again. As you know, you have more than just my sympathies. We have people who are lined up with baseball bats in hand to give our dual 2466's a lick if/when we can finally afford to move them out. In fact, one person I know is preparing a small but powerful explosive device to use on the whole pile...;-) > So rather than keep trying to fix these monsters I'm starting > to think about the cheapest way to keep the cluster running by > replacing just the mobo/CPU with something else (as I'm not > expecting enough $$$ anytime soon to do more, and obtaining > Athlon MPs and S2466N mobos now is problematical anyway.) > I'll happily give up Tyan's serial line bios access for a system > where I don't have to employ that feature quite so often! > > The S2466N is an ATX form factor, each one has one Athlon MP > 2200+ and 1 Gb of 2100 DDR RAM, a 40G ATA disk, a floppy > and a little PCI graphics card in a 2U case. If I could > find a nice mobo/CPU combo for, oh, <$200 that could > replace the S2466N and Athlon MP, and still do ECC, then I'd > probably go that route to patch systems up as they break. > Best if it has at least as much cache as the MP though. > Is there anything out there fitting > that description? Historically ECC support isn't something > that shows up on cheap mobos but maybe on some low end > Athlon 64 variant? I just bought an intermediate AMD64 mobo for my home cluster three days ago. The motherboard (ASUS) was $155, the CPU was $245 (for a slot 754 3400, which is really 2400 MHz, 128 MB L1, 512 MB L2 cache according to their numbering scheme). Also populated with a gigabyte of PC 3200 non-ECC memory it was about $500, and I'm very interested in seeing it go head to head with my opterons. I have NOT installed it yet so I can't give you any speed reports. Looking over the other prices from that vendor (intrex.com, although I go to a local store) their cheapest AMD64 is the 2800 for $150, and they had an MSI socket 754 motherboard for $100. All these motherboards seem to want non-ECC DDR400 (or slower) memory -- to get maximum performance you'd likely want to replace the memory anyway. So the absolute minimum sounds like it would be around $250, with a gig of new non-ECC memory around $350. Full speed ECC memory adds a LOT to this. Going socket 939 adds almost nothing (and I'm wondering if I should have gone this way after Bill's review of the AMD64 configs a few days ago). Going socket 940 adds a ton of money -- too much for a home box (and probably too much for your upgrade). Note that these are not pricewatch prices, so they are probably 20% or so more than you could find on the street if you tried hard. rgb -- Robert G. Brown http://www.phy.duke.edu/~rgb/ Duke University Dept. of Physics, Box 90305 Durham, N.C. 27708-0305 Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb at phy.duke.edu
- Previous message: [Beowulf] Keeping the Athlon MP cluster limping along
- Next message: [Beowulf] Keeping the Athlon MP cluster limping along
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
